d 
s 
lowing for the probable errors. Ten ground control points are 
uniformly identified in a Kanazawa area, whose location is 
shown in Table 2 and Figure 2. 
Table 2. Ground control points in Kanazawa area 
No . 
Geographic 
East 
longitude 
Deg.Min.Sec 
coordinates 
North 
latitude 
. Deg.Min.Sec 
Image coordinates 
Column Line 
indices indices 
. Sample Line 
Location 
1 
136.39.29 
З6.38.45 
177 
49 
Reservoir 
2 
136.40.27 
З6.З7.58 
200 
67 
Race track 
3 
136.36.36 
З6.37.О6 
112 
87 
Petroleum hase 
4 
136.37.19 
З6.36.32 
127 
100 
Estuary of Ono 
river 
5 
136.35.23 
З6.З5.56 
83 
n4 
Estuary of 
Kanazawa harbor 
6 
1З6.З9.ЗО 
З6.З6.ЗО 
178 
102 
Kanakusare 
river 
7 
136.43.ОЗ 
36.36.41 
259 
98 
Fukaya town 
8 
136.39.45 
36.33.49 
184 
164 
Kanazawa Uni 
versity 
9 
136.34.11 
36.31.49 
56 
209 
National rail 
way factory 
10 
136.40.56 
36.31.04 
211 
227 
Filtrat ion 
plant 
It should be mentioned that the residuals of the above GCP 
(u,v) were found within one pixel. It is of interest to me 
ntion that the geometric correction algorithms have recently 
been discussed by a few authors (cf.Bernstein 1976;Murai 
and Maeda 197 8) . 
is in digital form, it is required in the im 
age processing that the output lattice may be equispaced in 
the output space, because of the non-integral line and sam 
ple coordinates in the input space of the output picture ele 
ment. Then, it is necessary to interpolate over input space 
intensity values to generate each output pixel. This process 
of assessment is denoted as resampling,i.e., determination 
of the intensity value to be assigned to each output pixel. 
The geometric location of an output pixel in the input space 
is obtained by a ground control point method. Three methods 
of assigning the intensity value to each output pixel are 
Nearest Neighbor (NN), Bilinear Interpolation (BI), and Cu 
bic Convolution (CC). NN method finds the input pixel clo 
sest to the output pixel location and assignes that intensity 
value to the output pixel sample 
I (x,y) = I (u,v) , (5) 
where I(u,v) is the input signal being reconstructed. BI me 
thod uses four neighboring input values to compute the out 
put intensity by two-dimensional interpolation 
I(x,y)=3^1(u,v+l)+a 2 I(u,v)+a 3 l(u+l.v+1)+a^I(u+l,v), (6) 
where a^'s (i=l,2,3, and 4) are computed in advance. CC me- 
Resampling 
When the i